The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses the $ARGUMENTS variable to locate descriptor files. If the user provides a path-traversal string or a filename designed to trigger specific agent behaviors, it could bypass intended constraints.
[COMMAND_EXECUTION]: The skill allows the Bash tool. While the instructions focus on auditing, the audit includes checking install sections and CLI arguments within YAML files. If the agent attempts to validate these by executing them, it poses a risk of arbitrary command execution.
[EXTERNAL_DOWNLOADS]: The skill uses WebSearch and WebFetch to verify documentation URLs, IDE extensions, and tool capabilities (like SARIF support). This involves fetching content from arbitrary external websites based on the content of the audited YAML files.
[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection (Category 8).
Ingestion points: The skill reads and audits descriptor YAML files located in megalinter/descriptors/ or specified via $ARGUMENTS (SKILL.md).
Boundary markers: None. The skill does not provide delimiters or instructions to ignore embedded commands within the audited files.
Capability inventory: The skill has access to Read, Grep, Glob, Bash, WebSearch, and WebFetch (SKILL.md).
Sanitization: None. The content of the audited YAML files is processed directly to check properties like install scripts and linter_url without escaping or validation.

review-descriptor